I got an interesting request

I wonder if you would be willing to write a blog post looking at my recent test results in comparison to last year? Confirmed diagnosis of ME/CFS. UK NHS only helps with pacing advice.

• First BiomeSight test: 2025-04-17

Following this, 3 self-directed cycles of antibiotics, probiotics, prebiotics, and diet changes based on MicrobiomePrescription results. First 2 cycles increased my baseline and reduced symptoms dramatically, third cycle set me back slightly. Overall very positive.

Unfortunately then was hospitalised later in 2025 with a perforated and infected gallbladder, sepsis. They rotated through quite a few different harsh antibiotics trying to find one which worked. Then in December 2025 went in for surgery to remove the gallbladder, more antibiotics.

• Second BiomeSight test: 2026-03-16

My baseline now is worse again, many symptoms returned. I am loathe to use more antibiotics while some of my bacteria are so low (Akkermansia at 0.006) even though my positive scores are dominated by antibiotic suggestions. Would like to focus on probiotics, prebiotics, herbs, supplements, diet changes for now.

Any insight would be most appreciated.

Confirming the Worst

Going over to the symptom compare tool, we see that you are now worse than a year ago. 140 of 141 symptom forecasts are significantly worse! Seeing number this much worse is unusual but consistent with his events and perception.

Going Forward

My last two post has been evaluating the alternative path — instead of attacking the bacteria causing symptoms, push the person to a statistically significant healthy microbiome. The following links may be worth a reading:

• Another ME/CFS Microbiome Update – Review of a sample
• A new sample and two roads to walk – Review of a sample
• Good vs Bad Bacteria Dogmatic Beliefs – background
• Turning Fixing the Microbiome Upside Down! – background

This approach matches “I am loathe to use more antibiotics” because antibiotics typically are on the avoid lists with the healthy approach and high on the to take list attacking symptoms. It is sitting on the Simple UI page.

Basic Results:

• 52 bacteria were identified — every single one was too high.
• I have broken suggestions into classes below. In general, I have kept them to items with an impact of at least 1.
  • Items listed are order by largest impact first.

Herbs

The top herbs are below. I was delighted {Bofutsushosan} was listed because it is well known increases Akkermansia which he is concerned about.

Food

Flavonoids

Vitamins

Common and OTC Supplements

Probiotics x PubMed

This list is done using PubMed studies.

Probiotics x R2 Model

I prefer the R2 Model because we have a lot more data to use than with PubMed. On the flip side, this does not have clinical studies supporting the choices.

The top probiotic Bacillus thuringiensis suitable for human consumption may be a challenge. Most retail products are formulated to control caterpillars, worms, or mosquito larvae in gardens and standing water, not for ingestion or probiotic use.

Non-Human Retail options

• Home Depot sells Monterey B.T. Bacillus Thuringiensis Pint Concentrate as an online retail product.
• Wilco Farm Stores / Farm Store lists Bonide Bt Bacillus Thuringiensis as a garden pest-control product.
• DoMyOwn sells a dedicated Bt category and says it’s available through their store rather than big-box shelves.
• FBN lists Bt ingredient-based products, including Bacillus thuringiensis subspecies tenebrionis.
• DIY Pest Control lists Bt products and notes common trade names like Thuricide and Mosquito Dunks.

Summary

I look forward to see how well this alternative approach performs. It does not focus on the bacteria associated with his 141 symptoms — instead, we focus on shifting to a healthy microbiome profile (with very high statistical significance, p < 0.0001,) I would suggest retesting every 3-4 months to track progress.

Questions And Answers

Q: It’s interesting to see how some Odds Ratio based suggestions match with the Consensus Suggestions, and some vary wildly.

• A: Suggestions are based on bacteria target and available literature. Literature is sparse and often without replication of results
  • The safest path would be to start with items that are in agreement.

Q: I had one question with regard to whole milk, dairy, and lactose. The Odds Ratio analysis suggested these were positives – this makes my life a lot easier as I was using milk to help ferment and increase the CFU of the probiotics I used last year and hoped to again, and I eat a fair amount of dairy in general (I mostly eat a vegetarian diet with occasional fish, and dairy helps with my protein intake).

However when I ran the Consensus Suggestions earlier this week I got scores of -294.9 for bovine milk products, -120 for whole cow milk, and -158.6 for lactose.

• A: I favor the Odds Ratio. On this point as you have no issues with dairy, keep to your current usage.

Q: Does this mean I likely need to make a choice between the Consensus Suggestions route (which I followed last year) and the new Odds Ratio route?

• A: No, you could start doing a consensus of the consensus and odds ratio. Then add in items that disagree. I would suggest using an ratio evaluation:
  • Consensus: -120 with min of -960, so -(120/960) = -12.5%
  • Odds Ratio: 4 with max of 8, so 4/8 =50%
  • Thus odds ratio would win

Postscript – and Reminder

Other posts dealing with ME/CFS or Long COVID.

I am not a licensed medical professional, and the laws where I live prohibit any activity that could be interpreted as practicing medicine or giving personal medical advice. My work is limited to academic and analytical models, and I restrict myself to the language of science and statistics rather than clinical recommendations.

I cannot tell anyone what they should or should not take. Instead, I can present information about items that, based on numerical and statistical analysis, appear to have better odds of improving microbiome-related measures. I am a trained, experienced statistician with appropriate degrees and professional affiliations, and my role is to interpret data—not to treat patients.

All information I provide is for educational and informational purposes only and is not a substitute for professional medical advice, diagnosis, or treatment. Any ideas, rankings, or “suggestions” derived from my analyses must be reviewed and approved by your qualified medical professional before you decide to act on them.

The answers and explanations I provide describe my reasoning and methodology. They are not intended as medical advice for you or for anyone else, and they do not create a doctor–patient or provider–patient relationship. Always consult a knowledgeable licensed healthcare professional before starting, changing, or stopping any treatment, supplement, or health-related regimen.

This is the second review I have done since working through the implications of Mathematically Derived Healthy Microbiome. It highlights two very different strategies for improving gut health.

The earlier approach focuses on targeting bacteria that show statistical associations with symptoms. This approach often places many antibiotics near the top of the recommendations. When antibiotics are involved, I tend to favor the Cecile Jadin Protocol for ME/CFS.

The newer approach uses the revised model to target bacteria associated with a healthy, asymptomatic gut. In this approach, antibiotics often appear among the major items to avoid.

Both approaches are based on statistics, but the newer one has a much stronger statistical association.

The earlier approach has a track record of significantly improving the microbiome during the first few cycles. For some people, however, those improvements eventually stall. It also requires a friendly MD to prescribe the antibiotics, which is often a challenge.

If you apply the earlier approach one symptom at a time, the recommendations often contradict each other. “No man may serve two masters” becomes “No recommendations may heal two symptoms.”

The newer approach has no track record yet. It has only recently become available, and it is now being tried by someone whose progress has stalled.

So which one should you use? If you have a friendly MD, I would go with the earlier approach. If you do not, I would go with the newer approach.

Back Story

My symptoms have been somewhat confusing, but for many years ,like over 10 years ago I was constantly bloated with excessive wind/gas but also a lot of belching too. I would eat lots of wheat and sugar and processed foods. 2011 ended up on Proton-pump inhibitors (PPIs) on and off for over 10 years. 2016 appendix burst & got severe peritonitis and ended up very poorly, had 2 weeks of intravenous antibiotics. Slowly recovered.

Years of migraines and brain fog – but yet very active and social.

Then in 2022 I developed throat irritation that was exacerbated (i now believe by certain foods/Ingredients, alcohol, occasional smoking Definately fatty foods but I still cant quite put my finger on what made my throat irritation/ hoarse voice worse). I then developed Biliary Gastritis in October 2024 (Stomach lining erosion) likely from a possible intolerance just like the throat irritation. I became very constipated and still struggle with that.

• Foods that make me worse maybe
• Overly fatty foods
• Possibly milk
• Wheat, bread
• Possibly some fruit 
• Big blood sugar spikes off things like carrots and oats, sweet potatoes, very sensitive to carbs and sugar (i wore a glucose monitor out of curiosity) I am not diabetic.

Initial Review

There are two distinct paths, or algorithms, available in Microbiome Prescription.

The prior approach, which I call Traditional, begins with a few straightforward questions:

• Are you prepared to risk a severe Jarisch-Herxheimer reaction? It happens often.
• If you are working, can you afford to miss a few weeks?
• Do you have a friendly MD who is willing to prescribe a single course of each antibiotic listed below and become familiar with Jadin’s protocol?

The newer approach, Healthy Target, was recently added based on an odds-ratio model derived from healthy people. Instead of chasing symptoms, it shifts the goal toward a healthy gut. It does not address individual symptoms directly.

Building Suggestions

Probiotics

We actually have three ways of getting probiotics suggestions:

• Traditional Approach
• Healthy Target using Clinical Studies from PubMed
  • Data is very sparse on impacts
• Healthy Target using the “R2 Model” (a statistical model, not clinical studies)
  • Data is rich on impacts

We are going to compare only the positive probiotics from the R2 model that are easily available and the top 3 of the other models.

Bacteria	R2 Model	Healthy x PubMed	Traditional x PubMed
Streptococcus thermophilus	2534	4.4	-179
Enterococcus faecalis (Symbioflor-1)	1412	2.81	-262
Bifidobacterium infantis	1123	-2.77	-322
Bifidobacterium longum	762	0.74	-240
Bifidobacterium breve	389	-1.2	-204
Limosilactobacillus fermentum	47	1.31	60
Aspergillus oryzae	30	-1.34	-42
pediococcus acidilactici {P acidilactici}	n/a	7.8	75
lactobacillus paragasseri {L. Paragasseri}	n/a	7.8	-135
Escherichia coli Nissle 1917 {Mutaflor}	n/a	7.8	-12

Apart from Lactobacillus fermentum we have disagreement on positive or negative impact. A similar result is often seen when doing symptom by symptom with the traditional approach. We lack sufficient data to have certainty. Being a statistician, I favor the approach with the highest statistical significance — i.e. the novel or Healthy Target approach.

My Current Preference

There is nothing stopping a person trying one approach for 6-12 weeks and then retest; then switch to the other for 6-12 weeks; retest. Make sure that you keep detail notes on responses.

Sample Comparison Tool (example below)

• comparing the Healthy Microbiome Estimate from the two sample

Retrospective: Looking at what they reacted to

Food	Healthy x PubMed	Traditional x PubMed
Overly fatty foods	-1.88	262
Possibly milk	– 0.34	220
Wheat, bread	-1.36 (refined wheat breads)	-149 (refined wheat breads)
Possibly some fruit	1.73	-780.1
Range:	-7.82 to 7.82	-977 to 628
Focus area	-7.82 to – 3.91 3.91 to 7.82	-977 to -488 314 to 628

My impression is that the novel algorithm agrees better with their reactions. This shifts me further towards advocating for the novel algorithm.

Diet Plan

Often people want to simplify suggestions to one specific type of popular diet. This approach often defeat suggestions. Where diet are mentioned, they are secondary or tertiary guidance. Some generic diet studies appears in the suggestions, for example:

• restricted-fiber diet {low fiber diet} at 5.55
• Plant-based diet {Vegetarian diet} 4.57
• low-fat diets 4.37
• low fodmap diet 3.88
• ketogenic diet 3.19
• high-fat diets -1.88
• high-saturated fat diet -2.56
• methionine-choline-deficient diet {methionine-choline deprivation” diet} -2.82

This is intended as supplemental information to refine other suggestions where there is not sufficient information. The diets are what is cited in the literature. Most diets tend to be poorly defined. The classic example is Mediterranean diet. Often the “US Version” fails on the seafood or lamb aspects.

The exact foods vary by region: Greek-style diets may include more yogurt, feta, olives, and seafood, while other Mediterranean areas may use more pasta, beans, lamb, or different local vegetables and herbs. Even the meal pattern can differ, but the overall theme remains plant-forward, minimally processed, and olive-oil based.

Another example is the low-fiber diet. It is usually defined to be under 10 to 15 grams of fiber per day (about 1/2 of the recommended amount of fiber). However, studies show that the US population averages 15-16 grams per day!! so most Americans are already on a low-fiber diet

From the recommendations given I would build a general food diet from:

• 1 cup of blackberries each day, some lingonberry if available at a reasonable price
• Chicken as proteins source, no fish, little meat, no rare beef
• Pumpkin, banana, mushrooms, orka, carrots,
• Figs, banana, Lychee
• Citric juices
• pseudo-cereals {amaranth,quinoa, taro,buckwheat} – for example as a regular morning porridge

Avoid

• Fish
• Dates, Pineapple, nuts
• Celery, Asparagus, mustard greens
• refined wheat breads

The rest are herbs, spices and probiotics. Some avoids that may often be used with foods include:

• Cinnamon, Peppermint, Anise, Clove, Fennel, Sage, Basil, Garlic, Black Pepper, Table Salt

Suggestions are not indented to be dogmatic, rather shift towards one group of foods and seasoning and away from a different group.

Postscript – and Reminder

I am not a licensed medical professional, and the laws where I live prohibit any activity that could be interpreted as practicing medicine or giving personal medical advice. My work is limited to academic and analytical models, and I restrict myself to the language of science and statistics rather than clinical recommendations.

I cannot tell anyone what they should or should not take. Instead, I can present information about items that, based on numerical and statistical analysis, appear to have better odds of improving microbiome-related measures. I am a trained, experienced statistician with appropriate degrees and professional affiliations, and my role is to interpret data—not to treat patients.

All information I provide is for educational and informational purposes only and is not a substitute for professional medical advice, diagnosis, or treatment. Any ideas, rankings, or “suggestions” derived from my analyses must be reviewed and approved by your qualified medical professional before you decide to act on them.

The answers and explanations I provide describe my reasoning and methodology. They are not intended as medical advice for you or for anyone else, and they do not create a doctor–patient or provider–patient relationship. Always consult a knowledgeable licensed healthcare professional before starting, changing, or stopping any treatment, supplement, or health-related regimen.

I was messaged below:

Hello, could you tell me which antimicrobials are okay to use without killing the good bacteria? I have hydrogen SIBO, methane SIBO, and hydrogen sulfide SIBO. I don’t want to make things worse because I no longer have bifidobacteria, lactobacilli, and Oxalobacter in particular. And I don’t want to take something broad-spectrum.

I was especially wondering about clove and thyme. I also have fungal issues and yeast in my body, possibly related to mold. Could you explain how to tell whether an antimicrobial is harmful to the good bacteria? Thank you

What is defined as good or bad?

The issue is not that simple as “good” or “bad”. Too much of a “good” bacteria is associated with a variety of conditions. Let us look at the research for two commonly believed “good” bacteria:

• Lactobacillus is reported HIGH (from 119 studies) with
  • Allergies
  • Alzheimer’s disease
  • Amyotrophic lateral sclerosis (ALS) Motor Neuron
  • Asthma
  • Atherosclerosis
  • Atrial fibrillation
  • Autism
  • Carcinoma
  • Celiac Disease
  • Chronic Obstructive Pulmonary Disease (COPD)
  • Cognitive Function
  • Colorectal Cancer
  • COVID-19
  • Crohn’s Disease
  • Depression
  • Endometriosis
  • Fibromyalgia
  • Graves’ disease
  • Heart Failure
  • High Histamine/low DAO
  • hypertension (High Blood Pressure
  • Hypoxia
  • Inflammatory Bowel Disease
  • Insomnia
  • Irritable Bowel Syndrome
  • ischemic stroke
  • Juvenile idiopathic arthritis
  • Liver Cirrhosis
  • Long COVID
  • Metabolic Syndrome
  • Mood Disorders
  • multiple chemical sensitivity [MCS]
  • Multiple Sclerosis
  • Multiple system atrophy (MSA)
  • Neuropathy (all types)
  • Nonalcoholic Fatty Liver Disease (nafld) Nonalcoholic
  • Obesity
  • Osteoporosis
  • Parkinson’s Disease
  • Polycystic ovary syndrome
  • primary biliary cholangitis
  • Primary sclerosing cholangitis
  • Psoriasis
  • rheumatoid arthritis (RA),Spondyloarthritis (SpA)
  • Schizophrenia
  • Small Intestinal Bacterial Overgrowth (SIBO)
  • Stress / posttraumatic stress disorder
  • Systemic Lupus Erythematosus
  • Type 1 Diabetes
  • Type 2 Diabetes
  • Ulcerative colitis
  • Unhealthy Ageing
• Bifidobacterium is reported HIGH (from 120 studies) with
  • ADHD
  • Allergic Rhinitis (Hay Fever)
  • Allergies
  • Alzheimer’s disease
  • Amyotrophic lateral sclerosis (ALS) Motor Neuron
  • Ankylosing spondylitis
  • Anorexia Nervosa
  • Asthma
  • Atherosclerosis
  • Atrial fibrillation
  • Autism
  • Autoimmune Disease
  • Brain Trauma
  • Breast Cancer
  • Carcinoma
  • Cerebral Palsy
  • Cognitive Function
  • Colorectal Cancer
  • COVID-19
  • Crohn’s Disease
  • Depression
  • Endometriosis
  • Epilepsy
  • Fibromyalgia
  • Functional constipation / chronic idiopathic constipation
  • Graves’ disease
  • Hashimoto’s thyroiditis
  • Hyperlipidemia (High Blood Fats)
  • hypertension (High Blood Pressure
  • Hypoxia
  • Inflammatory Bowel Disease
  • Irritable Bowel Syndrome
  • ischemic stroke
  • Juvenile idiopathic arthritis
  • Liver Cirrhosis
  • Long COVID
  • Metabolic Syndrome
  • Mood Disorders
  • Multiple Sclerosis
  • NonCeliac Gluten Sensitivity
  • Obesity
  • Parkinson’s Disease
  • Polycystic ovary syndrome
  • Premenstrual dysphoric disorder
  • Psoriasis
  • rheumatoid arthritis (RA),Spondyloarthritis (SpA)
  • Rosacea
  • Schizophrenia
  • Sleep Apnea
  • Stress / posttraumatic stress disorder
  • Type 2 Diabetes
  • Ulcerative colitis
  • Vitiligo

There are a very small number of bacteria deemed absolutely bad.

• Bacillus anthracis.
• Francisella tularensis.
• Clostridium botulinum 

From Dangerous Microbes, 2018

The Human Need for Simplicity versus Biological Reality

I am a high functioning autistic spectrum individual. Others in the spectrum include those with photographic memory and complete memory recall. I lack those, but where I excel is my tolerance for complexity and uncertainty.

Across my 50-year career in software development, I’ve noticed that code I find straightforward often overwhelms other developers. One once remarked, “Any JavaScript file over 200 lines is black magic to me,” while reviewing what I considered a simple application. That experience reflects something broader: people naturally seek simplicity, even when reality is irreducibly complex.

In the same way, many approach microbiome science by labeling bacteria as “good” or “bad.” This reduction helps those who feel saturated by excessive detail—but the truth is far more nuanced.

The Evolution of Microbiome Prescription

For more than a decade, my goal with the Microbiome Prescription project has been simple in principle:

1. Accept scientific evidence—a microbiome test.
2. Compute suggestions aimed at correcting dysbiosis.
3. Provide direct links to supporting literature.

(See an example for Depression.)

The biggest challenge lies in determining which bacteria should shift, and in what direction. My early approach relied on lab-provided ranges: if a value was above range, reduce it; if below, increase it. But this method failed. Lab ranges are based on naïve averages and assume normal distributions. After teaching Ph.D.-level statistics, I knew better—bacterial populations follow heavily skewed distributions, not bell curves.

The next phase was to use symptom-annotated samples to mathematically model bacterial associations. When a new sample arrived, the system forecasted likely symptoms. Users checked which symptoms applied, improving both the model and predictive power.

Subsequent tests validated these forecasts: 53 predictions improved, while 19 worsened. It became clear that “gut health” cannot be captured by any single number. The ecosystem is too complex.

“No Protocol Can Serve Two Symptoms”

This phrase is an adaptation of Matthew 6:24: “No man can serve two masters.” When multiple symptoms are modeled independently, the results often conflict—what helps one symptom can worsen another. The earlier data illustrates this problem: 53 improvements, 19 regressions.

Rather than fighting symptoms individually, I began shifting focus toward the overall trajectory of health.

From Symptom Fighting to Health Trekking

A turning point came during an experiment using odds ratios derived from annotated microbiome samples—this time ranking bacteria by percentiles instead of percentages. Different labs report percentages inconsistently; percentiles normalize those variations (as discussed in this review).

Using 1,000 healthy individuals’ shotgun results from PrecisionBiome.EU, I noticed a striking pattern: “Asymptomatic: No Health Issues” consistently ranked as the top prediction.

That insight simplified everything. Instead of juggling countless symptom-specific models (10, 20, or even 200 symptoms), we can statistically track a single target—how far a sample deviates from “asymptomatic.” See definition here.

Now we’re just juggling one ball.

Reality vs. Model

The refined model depends on detailed microbiome tests—at least 16s sequencing, shotgun preferred—and percentile rankings for each bacterium. Unfortunately, most labs don’t provide percentile data. From Biomesight and Ombre, I can derive percentiles accurately from their percentage data. Some others attempt to estimate percentiles by assuming a bell curve—again, incorrect.

Recommendations for Individuals

Before ordering a microbiome test, confirm that it allows downloadable data with:

• Percentile and percentage values.
• Bacteria identified by NCBI Taxon numbers.

Recommended providers: Ombre or Biomesight (for better percentile reliability).

After testing, upload your results to Microbiome Prescription and simply click to start analysis.

Older analytical methods remain available and effective for many users, though progress may plateau for some. See Another ME/CFS Microbiome Update for details.

This is part of this continuing saga with this person. Prior posts and the labs shown below. Repairing the microbiome is not a single test, take a pill, and you are done. It may be like a long journey by sail through the fjords of Norway: a lot of course corrections!

• The ME/CFS Quest for Health [May 2025]
• Ongoing ME/CFS Journey [Jan 2025]
• ME/CFS Continues Improvement + Lab Read Quality Issues [Feb 2024]
• Rosacea, Circulation and mild CFS [Dec 2021]
• Update on ME/CFS Person [Sep 2023]
• Follow up Microbiome Analysis from a prior post [Apr 2022]

Person’s Summary

I would say that there is no improvement since the last test. So this is still applicable:

I have not been feeling so well lately (since the last year). I would say that my symptoms has become worse. Earlier it has always felt as I have done some progress but the last 18 months it has been the opposite. Earlier I got rid of my muscle and joint pain but it has come back and I have much bigger issues with my red nose and my body feels very stressed.

Also feel very bloated.

• A summary of my biggest issues:
• Get the red nose (some form of rosacea).
• Feel fatigued (both physically and mentally).
• Feeling stressed.
• Brain fog.
• Bloated.
• Lots of gas – I fart and burps a lot.
• Issues with allergies
• Muscle and joint pain

For the last 4-5 years I’ve been eating large amounts of rye and oats.

• Around 150-200 gram of rye bread every day.

• Around 70 gram of oats every day.

Been eating low fat, low protein and high carb (specially from rye, oats, apple juice and potatoes) because this diet seem to reduce my symptoms. As soon as I start to eat high meat and high fat my symptoms get worse.

In this analysis, I am going to look at:

• Changes since the last sample
• Review a new approach that is being incorporated
• Looking at suggestions and the difference between the new approach and the traditional approach

At the end, I suggest following the new approach with the motivation that the traditional approach has appeared to have stalled. The microbiome adapts to antibiotics and diet changes; rotation to alternatives often seems to be needed to keep destabilizing the microbiome dysbiosis.

Changes Between Samples

Going to Old UI/Multiple Samples we compare symptom matching values. We see that just 1 of 42 showed improvement.

Looking at the new Odds Ratio data, we see the number of bacteria identified as critical in different samples below / Odds Estimate. I am not clear on the meaning and significance…

• Odds Estimate: The higher the number, the more likely that the person is healthy
• Number of Bacteria: Not reliable because different bacteria contribute differently to health.

• 2026-03-06: 59 / 1632
• 2025-11-17: 36 / 1588
• 2025-03-30: 20 / 1671
• 2024-12-03: 21 / 1561
• 2024-09-02: 36 / 1611
• 2024-01-22: 58 / 1586
• 2023-09-12: 38 / 678
• 2023-02-22: 52 / 1707
• 2022-08-11: 30 / 886
• 2022-03-25: 24 / 1037
• 2021-12-03: 15 / 1287
• 2021-08-31: 49 / 757

My general reading is that from 2021-2024 there was improvement and now the person is in a new stable healthier state but with still dysbiosis. I am hoping that the switch to an alternative view of solving his health may result in further improvement. In other words, rotation of approaches.

Another View on the Same Issue

In my recent post, Turning Fixing the Microbiome Upside Down!, I introduced a different way to think about repairing the microbiome. A human–society analogy might make it clearer.

Imagine your city is struggling with homelessness, vagrancy, and petty crime. The usual response—especially in the U.S.—is to send in the police. Round up those panhandling on the streets! In microbiome terms, that’s like identifying “bad bacteria” and launching an attack.

But there’s another approach: offer housing, mental health care, and job training. You don’t punish people—you help them heal and reintegrate.

Traditionally, Microbiome Prescription has focused on detecting problematic bacteria and trying to suppress or adjust them. The challenge is that most people have many interconnected symptoms. Research often shows that substance X improves one symptom but worsens another. You end up chasing symptoms—fixing one only to see another emerge or intensify.

A more holistic alternative, which has only recently become possible, is to guide the person’s microbiome toward a naturally healthy state instead. See this post: Mathematically Derived Healthy Microbiome.

Recently I asked the head of a microbiome testing company, what statistical evidence do you have for what is a healthy or desired microbiome profile. How do you obtain the importance of each bacteria? He knew that using means and standard deviation were invalid because of the high skew with the data. His response was requesting his staff to remedy this situation, looking at odds-ratio as a starting point.

---

Evaluation

I am a modeler, not a medical professional. Modelers try putting together mathematics using available data and use that to generate predictions. Once the predictions are made, they are evaluated against any available facts.

Above we have some observations from the person, the model does not know this information — so we can evaluate predictions against this data.

Been eating low fat, low protein and high carb (specially from rye, oats, apple juice and potatoes) because this diet seem to reduce my symptoms. As soon as I start to eat high meat and high fat my symptoms get worse.

Comparing Suggestions

	Health	Symptoms	Novice
Rye	-0.11	-183.5	-352.3
Oats	0.15	-179.2	-742.6
Apple	2.34	-385.1	-351.5
Potato	1.32	-1102.7	-1418.5
High Meat	2.2	155.9	763
Raw Meat	-2.78	105.5	21.2
High Fat Diet	-.79	154.4	-246.1
high-saturated fat diet	-.35	96	-291.1
low-fat diet	0.86	195.9	378.4
Range of Values	-6.82 to 5.43	-2075 to 1359	-2296 to 1461
Best to Take	Catechol {Catecholamines}	restricted-fiber diet {low fiber diet}	restricted-fiber diet {low fiber diet}

Each of the above depends heavily on the bacteria selected and the threshold used. It is interesting to see that the new “Make Healthy” is a clear winner against his observations.

What is particularly interesting with the “Make Healthy” is that values were computed for 1,632 substances. Looking at the list os suggestions, we do not have a mass of antibiotics seen on the other lists. We are not focused on reducing bad bacteria, rather on improving the good bacteria, and letting those address the bad bacteria. The top items are below.

• Bromelain
• N-acetylneuraminic acid {Sialic acid} : a bioactive sugar involved in cell signaling, glycoproteins, glycolipids, and is abundant in the brain; it is linked to cognitive function, memory, and immune function in general mechanistic descriptions
• 1,2-dithiolane-3-pentanoic acid {a-Lipoic acid (ALA)}
• pediococcus acidilactici {P acidilactici}
• lactobacillus paragasseri {L. Paragasseri}
• Rubus fruticosus {Blackberry}
• Eutrema japonicum {wasabi}
• Cucurbita pepo {Pumpkin}
• Alkalihalobacillus clausii {Bacillus clausii }
• Faecalibacterium prausnitzii {Faecalibacterium prausnitzii}
• Capra hircus milk {Goat milk}
• 1,4-linked ß-D-xylopyranosyl {arabinoxylan}
• Gamma-Aminobutyric Acid {GABA}
• Ulmus rubra {slippery elm}
• Musa acuminata {Banana}

My impression is that this is a much friendlier set of suggestions. In fact, the bottom of the list (to avoid) are pages of antibiotics and prescription drugs.

Probiotics Exploration

There are two ways of getting probiotics:

• Using published studies on their impact. In general, each study describes one or two bacteria impacted. This results in low data
• Using the R2 Associations: This is a modelling of their impact with hundreds of bacteria impact estimated.

The new Healthy Algorithm includes R2 recommendations

We got the following suggestions

We will explore how different algorithms evaluate these.

Bacteria	Healthy With R2	Healthy With Studies	Symptoms With Studies	Novice with Studies
Streptococcus thermophilus	1988	-1 to 4	-132 to -139	-132 to -139
Enterococcus faecalis	1291	.6	-579	-579
Bifidobacterium infantis	1124	-2	76	76
Bifidobacterium breve	377	1	213	213
Bifidobacterium longum	365	1	-1423	-1423
Bacillus thuringiensis	96	n/a	n/a	n/a
Pichia kudriavzevii	38	n/a	n/a	n/a
Acidipropionibacterium acidipropionici	22	n/a	n/a	n/a
Aspergillus oryzae	15	.1 – 3	-192 to 8	-192 to 8
Lactobacillus acidophilus	4	0	-574	-574

Keeping to the “When in disagreement, leave it out” a.k.a. Minimal Risk a.k.a. “Do not harm”, we have

• Top choice is Bifidobacterium breve
• Reasonable choice is Bifidobacterium infantis

Alternatively, Streptococcus thermophilus high value, cheap, and easy availability — it is a good candidate to try a 2-4 week experiment.

Personally, I would be tempted to try the following pattern (starting at a low dosage and increasing):

• 3 weeks of Streptococcus thermophilus (up to 10 BCFU)
• 2 weeks of Bifidobacterium breve (up to 20 BCFU)
• 2 weeks of Bifidobacterium infantis (up to 20 BCFU)

Bottom Line

My personal choice would be to go with the “Healthy Algorithm” for the following reasons:

• The traditional approach has appeared to stall, time for a change
• It is heavily based on very statistical significance over the entire scope of bacteria involved (i.e. dense data) but it has not been validated by clinical studies.
• I have always been unhappy about clinical studies because the data is:
  • Messy (typically in the context of one or another medical condition)
  • Small sample sizes
  • Low resolution to bacteria

I am also curious to see how well the “Healthy Algorithm” performs.

Postscript – and Reminder

I am not a licensed medical professional and there are strict laws where I live about “appearing to practice medicine”.  I am safe when it is “academic models” and I keep to the language of science, especially statistics. I am not safe when the explanations have possible overtones of advising a patient instead of presenting data to be evaluated by a medical professional before implementing.

I cannot tell people what they should take or not take. I can inform people items that have better odds of improving their microbiome as a results on numeric calculations. I am a trained experienced statistician with appropriate degrees and professional memberships. All suggestions should be reviewed by your medical professional before starting.

The answers above describe my logic and thinking and is not intended to give advice to this person or any one. Always review with your knowledgeable medical professional.